Bug 243180 - security/afl: fix build on powerpc64 elfv2
Summary: security/afl: fix build on powerpc64 elfv2
Status: New
Alias: None
Product: Ports & Packages
Classification: Unclassified
Component: Individual Port(s) (show other bugs)
Version: Latest
Hardware: powerpc Any
: --- Affects Only Me
Assignee: Tobias Kortkamp
Depends on:
Reported: 2020-01-08 09:29 UTC by Piotr Kubaj
Modified: 2020-01-09 09:10 UTC (History)
1 user (show)

See Also:
tobik: maintainer-feedback+

patch (745 bytes, patch)
2020-01-08 09:29 UTC, Piotr Kubaj
tobik: maintainer-approval-
Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Piotr Kubaj freebsd_committer 2020-01-08 09:29:33 UTC
Created attachment 210521 [details]

Since powerpc64 elfv2 uses LLVM, it can build this port.

make test passes ok.
Comment 1 Tobias Kortkamp freebsd_committer 2020-01-09 05:44:24 UTC
Comment on attachment 210521 [details]

I'd rather you hook things up like in afl++ for aarch64 and powerpc64,
i.e., stop installing useless programs on non-x86 and making LLVM
option mandatory.  I personally do not see the point though.  We
already have security/afl++ for powerpc64 and aarch64, afl is stuck
on LLVM 8, and afl++ is more active than afl.

Why does afl need -mabi=elfv2 but afl++ does not?
Comment 2 Piotr Kubaj freebsd_committer 2020-01-09 09:10:51 UTC
(In reply to Tobias Kortkamp from comment #1)
I don't have aarch64 to test on, so I'll leave it up to you :)

afl needs -mabi=elfv2 because it uses LLVM 8. LLVM 9 has patches that for elfv2 that are autoapplied when building on elfv2, so that it doesn't use elfv1.

This is done with:
at https://svnweb.freebsd.org/ports/head/devel/llvm90/Makefile?view=markup
The extra patches are simply https://svnweb.freebsd.org/ports/head/devel/llvm90/files/ppc64-elfv2/patch-lib_Target_PowerPC_PPCTargetMachine.cpp?view=markup and https://svnweb.freebsd.org/ports/head/devel/llvm90/files/ppc64-elfv2/patch-tools_clang_lib_Basic_Targets_PPC.h?view=markup

They just set up clang to build for elfv2. It can be done without those patches by passing -mabi=elfv2.

LLVM 8 doesn't have those patches but can still build for elfv2 after passing -mabi=elfv2.

If afl++ is the superior alternative to this port (with only pros and no cons), then I guess it's fine not to fix it, but then why keep this port in the tree? :)